W3C home > Mailing lists > Public > ietf-http-wg@w3.org > July to September 2011

OWS in httpbis-p1-messaging

From: Frank Ellermann <hmdmhdfmhdjmzdtjmzdtzktdkztdjz@gmail.com>
Date: Thu, 25 Aug 2011 21:46:15 +0200
Message-ID: <CAHhFybpmpfjNJua0Sw4Afyhp33_Z3j14F_-EMXt3HwV_oxftRA@mail.gmail.com>
To: ietf-http-wg@w3.org
Cc: julian.reschke@greenbytes.de
Hi, in an IETF LAST CALL a draft for a proposed standard managed
to copy <OWS> syntax "by value", and fortunately I had the right
idea where the source might be: httpbis-p1-messaging-15 (now 16).

In essence that draft claimed (in prose) that producers of <OWS>
SHOULD (upper-case) limit their efforts to a single SP.

There are two problems with that approach:  Apparently folks do
not like the STD 68 (ABNF) HTAB in WSP, and arguably they have
a point based on RFC 5198 "Net UTF-8" section 2 point 3.

[The arguing party would be me, and visibly in RFC 5198 I lost.]
You could simply shift HT to the <obs-...> part of <OWS>, e.g.,

OWS     = *[obs-wsp / SP]
obs-wsp = (CRLF WSP) / HTAB
CRLF    = <RFC 5234 Internet standard newline>
WSP     = SP / HTAB                   ; adopted from RFC 5234
HTAB    = <RFC 5234 horizontal tab>
SP      = <RFC 5234 space>

Inventing a new name <OWS> for the STD 68 "multi-folding" oddity
could be another issue, did you really want to get rid of *all*
foldings?  That there is no line length limit in HTTP is not the
same as "we want no folding at all".

I'd prefer to use the same approach in HTTP as in RFC 5322 with
its <FWS> and <obs-FWS>.  But if discouraging any folding and/or
HTAB is the goal you will obviously need some kind of <OWS>.

It is still a bad idea to use <OWS> elsewhere in new HTTP header
fields such as Origin: before httpbis-p1-messaging is approved.

-Frank
Received on Thursday, 25 August 2011 19:47:24 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 27 April 2012 06:51:47 GMT